JPH0727442B2 - Method for improving hit rate in data storage device hierarchical structure and apparatus therefor - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to maximizing a hit rate in a data storage device hierarchical structure. More specifically, the present invention relates to transferring the most frequently used files to the most frequently used volumes in an automated storage library.

[0002]

Modern computers require a host processor that includes one or more central processing units and memory facilities. The processor processes the data in the memory according to the supplied instructions. Therefore, the memory stores the data that the processor needs,
It must have the ability to transfer that data to the processor at a rate that allows the computer's overall operation. Therefore, the cost and performance of computer memory is critical to the commercial success of computer systems.

Many forms of computer memory are available because modern computers require large amounts of data storage capacity. A fast but expensive form of memory is the main memory, usually composed of microchips. Another form of memory that can be used is called a peripheral storage device, which is a direct access storage device (DAS).
D), a magnetic tape device, an optical recording device, and the like. These types of memory actually store data on the storage medium. Although these types of memory have a higher storage density and lower cost than main memory, they do not provide the performance that main memory provides. For example, the time required to properly position a tape or disk under the read / write mechanism of a drive falls short of the purely electronic high data transfer rates of main memory.

Storing all the data in a computer in one type of memory device is inefficient. Storing all data in main memory is too expensive, and storing all data in one peripheral memory reduces performance. Therefore, a typical computer system
It includes a main memory and one or more types of peripheral memory, arranged in a data memory hierarchy. The arrangement of the data storage hierarchy is tailored to the performance and cost requirements of the user. In this storage hierarchy, main storage is often referred to as primary data storage, the next hierarchical level of the hierarchy is often referred to as secondary data storage, and so on. In general, the highest hierarchical level in the hierarchy has the lowest storage density and the highest performance and cost. As you go down the hierarchy, storage density generally increases, performance generally decreases,
Costs are also generally lower. Transferring data between different hierarchy levels in the hierarchy when needed results in the lowest memory cost and highest performance. That is,
Store the data in main memory only while the processor expects the data to be needed. This storage hierarchy can take many forms, including any number of data storage devices or memory levels, and being able to transfer data directly between two different memory levels. I / O channels, controllers, or cache memory well known in the art are used to transfer the data.

Images may be included in technical drawings, financial insurance documents, and medical charts / records.
Until recently, it has not been possible to store image data in memory in a cost effective manner. Images can take many forms and therefore cannot be encoded into computer binary values 0 and 1 as easily and compactly as text. Technical drawings are usually stored on paper, microfilm, or microfiche,
When you need access to a drawing, you have to search by hand. The same applies to X-ray and other medical diagnostic images, bank checks used for transactions between financial institutions, insurance records, and fax document images. Therefore, despite the development of computers, it is estimated that most of the world's data is still stored on paper. File a paper document containing these image data,
The cost of storing and retrieving has increased to eel climbing. Keeping a room or warehouse full of documents that must be manually retrieved when access is needed is no longer acceptable. Optical scanners can now convert images into machine-readable form that can be stored in peripheral storage, but the required storage space for image data is
Much less than needed for paper documents, but still quite large. Most business applications require large numbers of disks or tapes. Therefore, automatic storage libraries have been developed to manage the storage of such disks or tapes.

The automatic storage device library is a magnetic tape,
Includes a plurality of storage cells for holding removable data storage media such as magnetic disks, optical disks, a robotic picker mechanism, and one or more peripheral storage media access devices. Each data storage medium can be housed in a cassette or cartridge housing for easy handling by the picker. The picker, in response to a command, moves a data storage medium between a storage cell and an internal peripheral storage medium access device without the help of a hand. An internal peripheral storage medium access device mounted with a storage medium is said to be "occupied." After mounting the data storage medium on the internal peripheral storage medium access device, data can be written to and read from the medium as long as the system requires. The data is stored on the storage medium in the form of one or more files. Each file is a logical data set. A file is "open" when it is reserved for access by a particular user and the storage medium on which it is mounted is mounted in a peripheral storage medium access device and is ready for access. Is considered. For example, in an optical disc library, a file is reserved for exclusive access and is open if the disc on which it is mounted is mounted in a drive and spinning. Peripheral storage media access devices with mounted storage media containing open files are considered "active" whether or not an actual electronic transfer is taking place. Peripheral storage media access devices are also active when mounted storage media is accessed under standard operating system commands that do not require the file to be open, such as reading a directory. . Active storage media are generally considered to be storage media that are within an active peripheral storage media access device.
Internal peripheral storage medium access devices and storage cells can be considered at different hierarchical levels of the data storage hierarchy. Moreover, data storage media within shelf storage (ie, not within storage cells and out of reach of a robotic picker without manual intervention) is a further hierarchical level of the data storage hierarchy. Can be considered to be.

The automatic storage library may also include one or more external peripheral storage media access devices. An external peripheral storage medium access device, unlike an internal peripheral storage medium access device,
Peripheral storage medium access device that cannot be accessed by the picker and must be manually loaded and unloaded. An external peripheral storage medium access device may be included in the library for the convenience of the library operator. Shelf storage media that require only brief access do not need to be inserted into the library, nor do they need to be picked by a picker to be mounted in one of the internal peripheral storage media access devices. The external peripheral storage medium access device can also be considered as one hierarchical level of the data storage hierarchical structure. In the present specification, the "peripheral storage medium access device" means only the internal peripheral storage medium access device, unless explicitly stated otherwise.

Several automatic storage libraries are known. IBM introduced the IBM 3850 Mass Storage Subsystem for storage and retrieval of magnetic tape modules in the 1970s. More recently, several companies have introduced automatic storage libraries for magnetic tape cartridges and optical disks. For example, U.S. Pat. Nos. 4,654,727 and 4,864,438.
No. 4,864,511, a magnetic tape cartridge library is disclosed. Examples of optical disc libraries are U.S. Pat. Nos. 4,271,489, 4,527,262, 4,614,474, 47,665.
No. 81 in each specification. The robotic picker mechanisms of these libraries each include one or more grippers capable of handling one data storage medium at a time. Nos. 4,271,489 and 452 mentioned above.
Nos. 7262, 4614474 disclose robotic pickers with only one gripper, 46547.
Nos. 27, 4864438, 4864511, and 4766581 disclose robotic pickers having a plurality of grippers. IBM also sells the IBM 9246 Optical Library Unit, which is a two gripper library.

[0009]

Although the automatic storage device library is evaluated for its large online storage capacity, its performance is also important. One measure of automatic storage library performance is "hit rate", which is the number of file accesses per storage medium mount. As the hit rate improves, the number of mount and demount operations the picker has to perform decreases. Therefore, it is desirable to maximize the hit rate of the automatic storage device library.

A primary object of the present invention is to improve the performance of data storage device hierarchical structures.

Another object of the invention is to improve the performance of automatic storage libraries.

Another object of the present invention is to maximize the hit rate of the automatic storage device library.

[0013]

The above and other objects of the present invention are achieved by recording the frequency of use of each file in an automatic storage device library and a rewritable storage medium. For each file, the library counts the number of times that file is accessed. The library counts the number of mounts for each storage medium. Periodically, usage statistics are consulted to determine the most frequently used files and storage media in the library. The most frequently used files are then transferred to the most frequently used storage media and will be used more frequently.
The less frequently used files are transferred to the less frequently used storage medium and are used less frequently.
The most frequently used storage media are most likely to be mounted, thus improving the hit rate. Therefore, the total number of mount and demount operations required to access the file is reduced.

The present invention also provides several other functions. The operator can specify the files to be included on the most frequently used storage media, expecting frequent access. Also, during initialization of the automatic storage library, the most frequently used storage media can be preferentially mounted within the peripheral storage media access device therein. Means may be provided for causing one of the most frequently used storage media to be remounted immediately after being demounted. An automatic storage library suitable for implementing the above functionality is disclosed below.

According to one aspect of the present invention, a first hierarchical level storage device and a second hierarchical level storage device are provided, wherein
The data stored in the storage devices of the hierarchical level and the second hierarchical level can be accessed as files and can be transferred between the storage devices of the first hierarchical level and the second hierarchical level as a volume having a plurality of files. A method for improving the hit rate in a data storage hierarchy is provided. This method is the same as the second
Transferring a volume from a tier level storage device to the first tier level storage device, accessing files in the first tier level storage device, and tracking the number of times each file is accessed.
Recording the number of times of transfer of a volume from the storage device of the second hierarchical level to the storage device of the first hierarchical level; determining a file with a high access frequency and a volume with a high transfer frequency; Transferring the file to a volume that is frequently transferred.

The storage device of the first hierarchical level has a storage capacity of X volumes, and the frequently accessed files are transferred to X or less frequently transferred volumes.

The data stored in the storage device of the first hierarchical level is the data in the data storage medium in the peripheral storage device data access device of the automatic storage device library, and is stored in the storage device of the second hierarchical level. The stored data is the data in the storage medium stored in the storage cell of the automatic storage device library.

In yet another aspect of the present invention, a data storage hierarchical structure is provided. The outer layer structure includes a storage device of a first hierarchical level and a storage device of a second hierarchical level in which data is stored in such a format that the data is accessible as a file and can be transferred to each other as a volume having a plurality of files. A storage device, means for transferring a volume from the storage device of the second hierarchical level to the storage device of the first hierarchical level, means for accessing a file in the storage device of the first hierarchical level, and the access means. Means for recording access of each file and transfer of each volume from the storage device of the second hierarchical level to the storage device of the first hierarchical level; and a file having a high access frequency and a transfer frequency in response to the tracing device. A means for determining a high volume, and a means for transferring a frequently accessed file to a frequently transferred volume. Including.

In yet another aspect, an automatic storage device library for operating a storage medium including at least one volume consisting of a plurality of files is provided. The library moves a plurality of peripheral storage medium access devices, a plurality of storage cells each capable of storing one of the storage media, and the storage medium between the access devices and the storage cells. Means for mounting the storage medium and mounting each volume in one of the access devices, and the number of times each file is accessed by the access device and each volume in one of the access devices. Means for counting the number of times mounted by the mounting means, means for determining a file having a high access frequency and a volume having a high mounting frequency in response to the counting means, and a volume having a high mounting frequency for a frequently accessed file And means for transferring to.

The peripheral storage medium access device has a total mountable capacity of X volumes, and the frequently accessed files are transferred to the frequently mounted X or less volumes.

The peripheral storage medium access device has a total mountable capacity of X volumes, and the library further includes means for mounting X or less volumes which are frequently mounted at initialization.

The library further includes means for allowing an operator to designate a particular file as the most frequently accessed file.

[0023]

In the drawings, like numbers indicate the same features and structural elements in the various figures. The automatic storage library of the present invention will be described as implemented in an optical disk library. 1 to 4 show various views of such an optical disc library. The library 1 includes a housing 2 that houses most of the working parts of the library, the housing 2 having front and rear door plates (not shown) for accessing the interior. The library 1 further includes a plurality of optical disc storage cells 3 and a plurality of internal optical discs.
Including drive 4. Each storage cell 3 can store one optical disk having data recorded on one side or both sides. The data stored on one side of the disc is called a "volume". In the preferred embodiment, library 1
A total of 144 in which two columns of 72 memory cells are arranged
3 storage cells 3 and up to 4 optical disk drives 4
including. The optical disc includes an optical recording layer of an ablative type, a phase change type, a magneto-optical type or any other type, and may be a read-only type, a write-once type, or a writable type as long as it is compatible with the optical disc drive 4. Moreover,
Optical discs can be recorded in spiral or concentric track patterns. The details of the recording format are not part of this invention and may be any known in the art. The robotic picker 5 includes a gripper 6, which can access optical discs in any of the storage cells 3 or drives 4, the optical discs being placed in a cartridge for easy handling by the gripper 6. , Its form factor is 13.3 cm (5.2
5 inches), but any size compatible with the drive 4 and gripper 6 may be used.

Although the front of housing 2 is not shown in FIG. 1, some portions of library 1 project from the front of housing 2 for operator access. Those parts are part of the console door 7, power indicator / switch 8, inlet / outlet slot 9, external optical disk drive 10, console 1.
1. Including all or part of the keyboard 12. The console door 7 can be opened out, as shown in FIG. 2, so that it can be accessed behind it when needed. The slot 9 is used to insert an optical disc into the library 1 or remove it from the library 1. A command is sent from the operator to the library 1 via the keyboard 12, and the picker 5 receives the optical disk inserted in the slot 9 and transfers it to the storage cell 3 or the optical disk drive 4, or the storage cell 3 or the optical disk drive. It is possible to remove the optical disc from the library 4, pass it to the slot 9, and remove it from the library 1.
Using the console 11, the operator uses the housing 2
It is possible to monitor and control certain operations of the library 1 without looking inside. External optical disk drive 10
Unlike the internal optical disk drive 4, it cannot be accessed by the gripper 6 and must be loaded and unloaded manually. The library 1 also includes an optical disk drive exhaust fan 14, an external disk drive exhaust fan 15, and a power supply 16.

When the power of the library 1 is turned on, the command received by the keyboard 12 is transferred to the system controller 17. In the preferred embodiment, the system controller 1
7 uses the OS / 2 operating system
It is a BM PS / 2 personal computer. The IBM PS / 2 personal computer includes main storage and one or more storage media, such as storage media in fixed disk or floppy disk drives.
The system control device 17 uses the optical disc
Drive 4, external disk drive 10, picker 5
Issue an order to. The drive control card 13 and the picker control card 18 use a known small computer system interface (SCSI) command bucket issued from the system control device 17 as an electrical signal for the optical disk drive 4, the external disk drive 10, and the picker 5. Convert to mechanical action. The movement of the picker 5 in the library 1 is in the XY direction. The vertical movement is driven by the vertical motor 19, and the horizontal movement is driven by the horizontal motor 20. The motor 19 rotates the lead screw 21 to move the picker 5 vertically, and the motor 20 rotates the belts 22 and 23 to move the picker 5 horizontally. Further, the picker 5 is rotated to move the gripper 6
It is also possible to place either surface of the optical disc within the reach of the upright position. Other physical features of the library 1 are not shown in the drawings for simplicity of illustration and are shown and not numbered, but are all known.

Next, the system connection of the library 1 will be described with reference to FIG. The system controller 17 is
One or more host / system processors 30,
It is connected to receive input from it and send output to it. The system processor 30 is a host central processing unit (CPU), such as an IBM 3090 mainframe processor using the MVS or VM operating system, an IBMS / 40 using the OS / 400 or AIX operating system.
0 Medium-scale computer, OS / 2 or DOS
IBM P using operating system and located in a local area network (LAN)
It may be a network of processors such as S / 2 personal computers. Although not shown, the connection to system processor 30 is well known. If the system processor 30 is an IBM 3090 mainframe processor, the IBM document "IBM Channel-to-Channel Ad"
apter "(Document number SA22-7091-00), June 1983, I
BM document "IBM System / 360 and System / 370 I / O Interfa
ce Channel to Control Unit, Original Equipment Manu
facturers Information "(Document number GA22-6974-09), 1
February 988, and IBM document "IBM System / 370 Exte
The connection can be made using the IBM System / 370 channel attachment mechanism based on the interface described in "Neded Architecture Principles of Operation" (Document No. SA22-7085-01), January 1987. System Processor 30 Is an IBM AS / 400 medium sized computer, a direct connection can be made using the SCSI interface attachment, in which case the library 1 is ANSI standard X3T9.2 / 86-109.
rev. 5 directly controlled by the host system. If the system processor 30 is multiple IBM PS / 2 personal computers located in a LAN, the IBM document "Distributed Data Mana"
gement Level2.0 Architecture Reference "(Document number S
C21-9526), March 1989, and a connection can be made using the NETBIOS control program interface of the IBM Token Ring Network LAN Attachment. In the following, a preferred embodiment of library 1 will be described in which a library 1 appears to the system as a shared general purpose storage device, a LAN.
It will be described as being used as a file server in the environment.

The system controller 17 uses the SCSI bus 3
1 is connected to the optical disk drive 4, the picker 5, and the external optical disk drive 10 via a known single-ended SCSI connection. In another embodiment, the system controller 17 is similarly placed in another physical box (not shown),
It can also be connected to direct the operation of the box. The box is essentially the same as that shown in FIGS. 1-4, except that the box does not physically contain the system controller therein, but is instead controlled by the system controller 17 via the SCSI bus 32. Is. A logical subsystem containing one box containing the system controller and one box not containing the system controller is considered as one library. In addition, for use in certain environments, two system controllers are connected via an RS-232 interface (not shown), including two boxes each containing a system controller and two boxes not including a system controller. You can also create a library, and so on.

Next, referring to FIG. 6, the system controller 1
A description will be given at the level of seven functional components. In general, the system controller 17 is responsible for creating and deleting files, writing to and reading from files, moving optical disks between storage cells 3 and optical disk drive 4 and slots 9, usage and error statistics. Offer, etc.
Designed to support major library features. A volume in the library represents a subdirectory name within the root directory of a single drive. The system processor 30 can read the root directory but cannot store files in the root directory. Any path accessed on a volume will appear as a path under the subdirectory element that represents that volume label.

General library file server (GL
The FS) 50 controls the library with a set of general purpose intermediate hardware commands via a well-defined interface. This interface will be described later. The data is handled by the GLFS 50 at the logical record level, allowing access to data amounts ranging from single bytes to entire variable length data objects. The operating system 51 arbitrates the flow of control and sends incoming operating system commands from the external interface to the library subsystem. Operating system 51 may be any of several well known operating systems,
In the preferred embodiment, it is the OS / 2 operating system. The OS / 2 operating system generally allows control of library 1 with standard fixed disk operating system commands. Control of the library is done by the original command DosF
Commanded by sCt1. Using this command, initialization, loading / unloading of an optical disk into the library 1, reading / writing of a library map file, mounting / demounting of an optical disk in the drive 10, etc. are supported. Drive control is commanded by a unique command DosDevIOCt1. The rest of the programmed control for library 1 is
It is held in microcode, which is uploaded from the storage medium to the main memory of the system controller 17 at initialization. In other embodiments, some of the functionality needed to support microprogrammed control may also be provided as a utility to the operating system running within system processor 30.

The OS / 2 operating system is
It includes several advanced operating system concepts incorporated into the system controller 17. This advanced concept includes dynamic link libraries, installable file systems, and multitasking. Dynamic link
A library (DLL) is a file that contains a set of functions that can be dynamically loaded as needed. Normally, a program must be compiled and linked with the compiled program code for all the functions that it may call before it can run. DLLs allow programs to call functions that have been compiled and linked into independent modules of program code. OS / 2 includes a set of DLL modules that can be called as needed. Custom DLL
Modules allow OS / 2 to control non-standard storage devices. The custom DLL module is called the installable file system (IFS). Each function supported by IFS is called an entry point. For more information on deployable file systems, see the IBM document "IBM Personal Systems Develope"
r "(Document No. G362-0001-03), Fall 1989. In the preferred embodiment, GLFS 50 is implemented as an IFS to an OS / 2 operating system with a defined entry point.

Another important aspect of the OS / 2 operating system is multitasking. Multitasking is the ability of a system to execute many programs simultaneously. System processor time
It is distributed among the tasks, and each task appears to be executing as if there were no other tasks. A separate environment is maintained for each task. The memory and register contents for each task are separated to avoid mutual interference. “Thread” a task and its associated environment
Call. The program may include a code area and a data area in the main memory of an IBM PS / 2 80 type personal computer. A code area is a section of memory that contains instructions that are executing for any given thread. A data area is a section of memory (or register) that is manipulated during execution of an instruction. Since the same code area may be used for multiple threads,
Each thread may refer to the same code area to be executed, but each thread contains a separate data area.

Upper interface translator 8
0 governs the exchange between the upper interface command and the GLFS 50 command. The lower interface translator 90 manages the conversion between the command issued by the GLFS 50 and the lower interface command. The translators 80 and 90 are implemented as separate linkable modules, each with a well-defined interface, to allow the library 1 to be easily connected to the new upper and lower interfaces. The only effect of the connection on the new interface is that new parts of translators 80 and 90 are created. GLF
Since S50 is generic, it remains unchanged.

The upper interface of the library 1 includes a library organization file, a library map file, a system performance file, a console 11 (and keyboard 12), and a network interface. Although not shown in the drawing, the library configuration file, library map file, and system performance file are stored in the fixed disk drive of the system controller 17. These files are maintained by the library operator or maintenance personnel. The library configuration file includes various characteristics of the hardware configuration of the library 1, such as the number of physical boxes in the library 1, the number of drives 4 and 10 in each physical box, and whether the drive is an internal drive 4 or an external drive 10. , The number of storage cells in each physical box, picker 5 and drives 4 and 10
List SCSI addresses, etc. The library map file contains various characteristics of the optical discs in library 1, such as the volume label of each optical disc in library 1, free space information for each optical disc in library 1, and some usage statistics for each optical disc. For example, the number of mounts, the date and time of the last access, the number of times each file on the optical disk is accessed, etc. are listed. The system controller 17 uses the library configuration file and the library map file to
Identify the number and sequence of resources in the library, library 1
As the status of resources within changes, the files are adjusted accordingly. The system performance file lists certain operator-specified parameters such as the frequently accessed file (FAF) specified by the operator and the maximum number of frequently mounted volumes (FMV). These parameters will be defined later. Consol 11 is
Used to indicate the progress of library components and to make command and utility functions such as error reporting available to the operator. Keyboard 12
Used by an operator to make manual inputs to the library 1 in response to information received, for example, via the console 11. The console 11 and the keyboard 12 are a console driver 81 and a console digital
Linked to GLFS 50 by manager 83. The network is a LAN adapter driver 82 and N
It is linked to the ETBIOS network control program 84. This network interface allows a processor on the network to gain remote access to the library 1 which acts as its file server.

The GLFS request manager 52 is an interface to the operating system 51, and
2 Responds to the same set of entry points that the operating system uses to communicate with IFS. The GLFS request manager 52 is responsible for splitting operating system commands to perform library functions, and does so by calling routines in the process control manager (PCM) 53a to perform each step. PCM 53a is a set of utility routines that assist the system in decomposing and processing commands, including those that require the generation of request blocks. These routines parse the directory path string, put the optical disk into the library, find the location of the volume, map the drive to the volume,
Suggest to flip the optical disc and mount the volume on the opposite side, mount the volume, demount, eject the optical disc from the library, and so on. These routines will be described later in this section. The directory management method (DMS) 53b is a code module that satisfies the IFS file designation for monitoring the open / close status of user files in the library 1 as is well known, and is for operating such user files. To use. The use of the IFS interface in such an internal module allows for easy adaptation of external IFS style directory management scheme implementations.

Power-on initialization (POI) module 5
Reference numeral 4 controls the power-on and reset functions of the control device, and the operating system 5 is used at the time of initial setting.
Called by 1. The POI module 54 is responsible for functions such as determining and reporting the results of component self-tests, reading library configuration files and status files. Errors are handled by error recovery module 56 and error recording module 57. The error recovery module 56 handles all errors, including dynamic device reallocation and device command retries. The error recording module 57 stores the error information and the console 1
It is responsible for reporting it to the operator via 1.

The resource manager 60 dynamically allocates and deallocates control blocks in the data area of the system controller 17, including the request block, drive control block, and error information block. The request block is used to request a hardware event for the optical disc drive 4 or the picker 5. The drive control block is used to store status information about the drive 4, as will be explained later. The error information block is used to store the information needed to perform error reporting, isolation, and possibly retry. Control block allocations and deallocations are performed using a list of available free space in main memory of the IBM PS / 2 personal computer maintained by resource manager 60. The error recovery module 56 and the resource manager 60 are both connected to most of the components of the system controller shown in FIG. These connections are not shown for clarity of illustration.

The schedulers 61 and 62 are responsible for verifying a portion of the contents of the request block and putting it into the pipe of the hardware device processing the request. A pipe is a wait data path from one thread to another thread that can be accessed by any thread that knows the identifier assigned to that pipe. The dispatchers 63 and 64 validate the request block to ensure that the request is ready to be executed,
It is responsible for appropriately assigning the request to the drive logic manager 91 and the library logic manager 92. The coordinator 65 controls the execution of requests to the dispatchers 63 and 64. The coordinator 65 is a PC
It does so using a table with entries for each request block received from M53a. Each entry lists the assistance request block associated with the particular request block. A request that requires another request to be completed is called a "dependent" request, and a request that must be completed first is called a "support" request. Coordinator 6
5 suppresses the execution of dependent requests until the associated support request is executed. If the request for assistance is not fulfilled, the coordinator 65 rejects requests that are subordinate to it.

The logical managers 91 and 92 are responsible for translating generic library commands in the form of request blocks into corresponding device level commands in the form of SCSI data packets. The logical managers 91 and 92 also have respective drive drivers 93 and library
It is responsible for receiving the hardware status information from the driver 94. Drivers 93 and 94 directly operate hardware and physical memory. Drivers 93 and 94 also
Carry out all communication with the respective hardware in question,
And respond to the interrupt. The logic manager 92 and library manager 94 control the picker 5. Although not shown in FIG. 6 for the sake of simplicity, in practice, a large number of drive dispatchers 63, drive logic managers 9
There is one drive driver 93 for each drive 4 or 10 in the library. Each set is connected to a different data pipe.

Operating Method Initialization of the library 1 is performed by operating system 51, GLFS request manager 52, resource manager 6
0, executed using the POI module 54. After verifying that each is working properly by self-checking the library hardware, the operating system 51 is loaded and the OS / 2 CONFIG. SY
The S (system configuration) file is used to set operating system parameters and load drivers. The operating system 51 then generates an initialization command, which is passed to the GLFS request manager 52 and then to the POI module 54.
The POI module 54 reads the library configuration file, the library map file, the system performance file, creates the necessary internal data structures in the main memory of the IBM PS / 2 personal computer, and specifies them in the library configuration file. Initialize a separate thread for each hardware component of the library. The resource manager 60 initializes internal tables used for memory management. Then POI module 5
4 is a system controller 17 and control cards 13 and 1
8. Query the result of the power-on self-test at 8 and report any problems to the error recovery module 56. The error detected during initialization is the error storage module 57.
Stored by the error recovery module 56 and possible. After error detection and logging, the POI module 54 sets the optical disk map 115.
Inquires to determine the FMV (most frequently mounted volume) in the library and then initiates a request to prioritize mounting of such most frequently mounted volume on drive 4. To do. When the system controller 17 is ready, the system is sensitive to the activity of the console 11 or network interface.

Referring to FIG. 7, the internal data structure is the optical library main system control block (OLMSCB) 1
10, 1 or more library control blocks (LC
B) 111, fixed disk control block (DCB) 11
2. Active DCB pointer array 113, active DCB1
14, an optical disk map 115, and a file usage database 116. The pointer is indicated by an arrow in FIG. The OLMSCB 110 includes the number of physical boxes in the library 1, the maximum number of FMVs, a pointer to the optical disk map, and a pointer to the LCB 111 for each physical box in the library 1 (for convenience,
Only one LCB is shown in the figure). Each LCB111
Regarding the physical box in question, the operating status of the picker 5 (online, offline, failure), the number of drives 4 and 10 therein, the SCSI address of the picker 5 therein, the number of storage cells 3 therein, It contains the address of each storage cell 3 and slot 9 therein and a pointer to the fixed DCB 112 for each drive 4 or 10 therein. Each LCB 111 also includes a pointer to the active DCB pointer array 113, and the pointer array 11
3 is the active DC for each drive 4 or 10 in it
Contains a pointer to B114.

In the figure, each drive 4 is shown in the preferred embodiment.
And 5 fixed DCBs 11, one for every 10
2 is shown. Each fixed DCB 112 is a drive 4
And some type 10 drive specific information. These pieces of information are “fixed” information because they do not change when the optical disc is operated with respect to the library 1. This information includes the operating status of the drive, including usage attributes indicating whether the drive's use is restricted to certain functions (such as write-only). The fixed DCB 112 is used as a permanent record of such information to create an active DCB 114 when an optical disc is operated around the library 1, as described below.

In the figure, a total of six active DCB pointers 113 and 114 are shown, one for each drive 4 and 10 and one for each virtual list in the preferred example. This virtual list, as explained later,
It is a linked list of access records for certain volumes. Active DCB 114 contains some volume-specific information about drives 4 and 10 and virtual access. This information is "activity" information as it changes (ie is updated) as the optical disc is operated around the library 1. This information is
Fixed DC for each drive or virtual access in question
Contains the appropriate information from B112, and also the occupancy status of the drive (whether the optical disc is mounted),
Usage statistics, such as the last access time and the user count for the volume or virtual access within it, an index into the optical map for the entries therein describing the volume or virtual access mounted in the drive, and file usage Contains an index into the database. The index to the optical disk map is used by the DMS 53b to find the position of the volume in the library 1 as needed. A user·
Count is the number of current ongoing accesses to a volume. Access is an open file or any standard operating system command that does not require opening a file, such as reading a directory.

The optical disk map 115 contains an entry for each storage cell 3 in the library 1. The entry for empty storage cell 3 is blank. Entries for a full storage cell 3 include the owner of the disk within it, the home storage cell 3 and current location of that disk, and for each volume on that disk, the volume label, number of mounts, and free space available. List space and other usage statistics. File usage database 1
16 is a library 1 designated by a volume
Contains an item for each file stored in. Each file item contains the size of the file, the number of times the file was accessed, and an indication of whether the file is an operator-specified FAF. The above data structure also includes other information not shown in the figure for simplicity of illustration, but which is necessary for the operation of library 1, as is well known in the art.

Next, referring to FIG. 8, the system controller 1
The operation of No. 7 will be described. Upon receiving a request from the network interface, the network control code
In step 100, the request is sent to a set of standard OS / 2
Translate into operating system commands. The operating system 51 then proceeds to step 101.
, Issue the appropriate operating system calls to process those operating system commands. The GLFS request manager 52 receives these system calls and breaks them down into simpler functions. For each function, the GLFS request manager 52
In step 102, a routine in PCM 53a and / or DMS 53b, or both, is called and the appropriate subset of data required for that routine is passed as a parameter. For each routine that requires hardware activity, PCM 53a and / or DMS 53b, or both, call the resource manager 60 to generate a hardware level request block at step 103,
It issues the block to schedulers 61 and 62 and informs coordinator 65 of any hardware dependencies so that the requests can be properly ordered.
PCM 53a also returns control and status information to GLFS request manager 52 as each routine completes.

After examining the list of free space available in the main storage of the IBM PS / 2 personal computer, resource manager 60 allocates the required memory space for the requested block. The routine that calls the resource manager 60 provides most of the information for a control block, and the resource manager 60 fills in some additional information such as the control block identifier and the request block identifier. The drive scheduler 61 and the library scheduler 62 receive all hardware event requests as request block identifiers and transfer them to the data pipes connected to the drive dispatcher 63 and the library dispatcher 64, respectively. Dispatcher 6
3 and 64 indicate whether the requested block identifier is present,
Inspect each data pipe sequentially. The coordinator 65 is called to determine whether the request block is ready for execution. Coordinator 6
5 examines the request block dependency table and dispatcher 6 until all support request blocks are complete.
Prevents 3 and 64 from issuing the request block identifier.
When all required block dependencies are satisfied, the required block identifier is issued to the relevant physical manager 91 or 92.

In step 104, the logical managers 91 and 9
2 receives the request block identifiers and the SCSI hardware commands needed to fulfill those requests.
It creates packets and issues them to drivers 93 and 94. The hardware then physically fulfills those requests. As each request completes, the logical managers 91 and 92 signal its completion. The dispatcher 63 or 64 then issues the identifier of the next request block to the relevant logical manager 91 or 92.

In general, library mount / demount operations are performed as long as there is a volume mount request.
Continue as needed. Drive 4 or 10 first
When a volume is mounted on the active volume, an active DCB 114 for access to that volume is created. The active DCB stores drive-specific information about the drive 4 or 10 on which the volume is mounted in a fixed D
It is created by copying from the CB 112 to a block in memory and adjusting the pointer appropriately. During access to the volume, volume-specific information about the access is updated and stored in the active DCB 114. When the volume is demounted, volume-specific information other than the occupancy information is deleted from the active DCB 114, indicating that the drive 4 or 10 in question is emptied again. When the volume is remounted on the drive 4 or 10 in question, the active DCB
114 is updated again with appropriate volume-specific information as needed, and so on.

When the volume is demounted, if drive 4 is inactive, drive 4 will be able to serve existing mount requests, thereby keeping drive 4 occupied. Such occupancy maximizes the amount of data that is immediately accessible. However, when there is no pending mount request, the drive 4 is mounted so as to ensure the existence of an unoccupied drive 4 in order to meet the future mount request and reduce the deterioration of the drive 4 during the idle period. It can be demounted in advance. If all drives 4 are occupied, one drive 4 can be emptied. In addition, the activity status of the drive 4 is periodically checked, and since the library 1 is relatively idle, one of the occupied drives 4
Determines if the volume within should be demounted in advance. Thus, during normal library operation, drive 4 will only be empty after a volume has been previously demounted. When all the drives 4 are occupied and there are no pending mount requests, the criteria for selecting the volume to be demounted in advance are:
It differs from the criteria used during the periodic check of drive 4 activity.

Drive 4 can only drive one at a given time
Only one optical disc can be physically written or read. When a volume mount request is issued when there is no unoccupied inactive drive 4, the request is normally rejected. However, using the virtual drive function, when all the drives 4 are occupied and active, a mount request for the volume temporarily demounts the least recently used volume, thereby causing the existing drive 4 Allows access to more volumes. Volumes that are temporarily demounted are said to have been "swapped out" and newly mounted volumes have been said to have been "swapped in". Drive-specific information about drive 4 is deleted from active DCB 114, while volume-specific access information about temporarily demounted volumes is stored therein. Then the active DCB 114 for the temporarily demounted volume.
Are held in a special form of active DCB 114 called a virtual list. The virtual list DCB differs from other active DCBs in that it contains pointers to each other, resulting in a linked list. By using this virtual list, it is possible to resume the operation on the volume by remounting it at the next volume mount request. It is also possible to use cache processing to continue access without a remount operation. Upon remounting the volume, the virtual list DCB in question is removed from the linked list and the volume specific information is copied to the active DCB 114 for the drive 4. Because such access information is retained, swapped out volumes are still considered active and are being accessed. Also, the swapped-out volume can be remounted in any drive 4 as long as the active DCB 114 for that drive is provided with access information. Volume access is not bound to the original drive 4 on which the volume was mounted. The swapped out volume does not appear logically in its home storage cell 3. This is because remounting must be distinguished from mounting a volume that was not swapped out. Therefore, the actual number of drives 4 in the library is transparent to the user.

The following explanation regarding each of the above routines will be given below.
For convenience, the previously described features and features well known in the art will be omitted and simplified if possible. For example, OLMSCB11
0, LCB111, DCB112 and 114, the optical disk map 115, and the information in the file use database 116 have already been described with respect to their position, content and use, and will not necessarily be mentioned. The term "return" refers to leaving a routine and going to the step that called the routine, including somehow directing the result of the routine to the step that called the routine.

Next, referring to FIG. 9, the system controller 1
Representative IFS to G6FS50 requesting access to a particular designated volume for high level 7 operations
We will start from the entry point and explain in detail. Step 200
GLFS to request access to the specified volume.
The request manager 52 receives from the operating system 51 and then moves to steps 201-210 to call various PCM 53a routines. Step 201
Then, call the PARSE routine to extract the volume label from the request and locate the specified volume using the optical map. At step 202, a branch is taken depending on the result of the PARSE routine. If the PARSE routine returns an error message (ie, did not complete successfully), then the error message is returned at step 210. PARS
If the E routine is successful, step 203 branches to the specified volume according to position. If the specified volume is not found in library 1, then it cannot be accessed. Therefore, flow jumps to step 210 and an error message is returned. The specified volume is library 1
If found in step 204, READY in step 204
The VOLUME routine is called. READY V
The OLUME routine is used to determine some later required actions depending on the position of the specified volume. Upon completion of the READY VOLUME routine, step 205 reads the READY VOLUME.
Branches according to the result of the E routine. READY VO
If the LUME routine returns an error message (ie, did not complete successfully), step 210.
Will return that error message. READY V
If the OLUME routine is successful, step 207
Then, the operation for the specified volume is performed according to the request. Once those operations are complete, step 208
Then, the RELEASE VOLUME routine is called in order to determine whether it is appropriate to demount the volume in advance even if the library 1 is not in the idle state. Upon completion of the RELEASE VOLUME routine, the result is returned in step 210.

In operation, the routine may call other routines. For example, RELEASE VOL
UME routines can call routines to mount, demount volumes, swap in and swap out using virtual lists, and so on. These additional routines, other than those described further herein, are not critical to the invention.

One routine that is sometimes called during library operation is the MOUNT routine. Referring to FIG. 10, the MOUNT routine begins with the request at step 120. In step 121, the picker 5 retrieves the specified volume from its current location. Next, in step 122, the designated volume is transferred to the allocated drive 4, and in step 123 it is inserted therein. In step 123, the allocated drive is also spun up to operating speed. At step 124, internal data structures are updated as needed, including incrementing the number of mounts for the volume in the optical disk map 115. The MOUNT routine returns at step 125.

The file size and file access number for each file in the file usage database 116 are updated in the same manner as the mount number on the optical disk 115. File mount database 11 for each file access after the volume is mounted
The file access count for that file in 6 is incremented. Such updating of the file usage database 116 is done in step 207, but is
No further description is given here.

Another routine that is sometimes called during library operation is the DEMOUNT routine.
Referring to FIG. 11, the DEMOUNT routine begins at step 130 with its request. In step 131,
The designated volume is decelerated in rotation and searched by the picker 5. Then, in step 132, the specified volume is transferred to its home storage cell 3 and is inserted therein in step 133. In an alternative embodiment, the designated volume may be transferred to another storage cell 3 (ie, not its home storage cell) if it is an FMV. For example, the one closest to drive 4
The set of storage cells 3 can be reserved exclusively as occupied by the FMV. In another example, the FMV can be transferred to the empty storage cell 3 closest to the drive 4. In these alternative embodiments, keeping the FMV as close as possible to drive 4 further improves library performance. At step 134, whether the demounted volume is an active non-FMV demounted to handle an existing mount request, or an FMV deactivated to handle an existing non-FMV mount request. State or inactive state). If the demounted volume falls into any of the above categories, then in step 135 an item is created in the virtual list for the demounted volume. Such creation of a virtual list item replaces the virtual list item. Non-FM with demounted volume inactive
If it is V or an active FMV that has been previously demounted, proceed to step 136, at which point the otherwise otherwise normal update of the internal data structures is done. The DEMOUNT routine proceeds to step 137.
Return with.

At any point during the operation of the library, the operator initiates a DESIGNATE routine,
The maximum number of FAF or FMV designated by the operator can be changed via the keyboard 12. Next, with reference to FIG. 12, the DESIGNATE routine will be described in order from the request in step 140. Step 14
At 1, the operator is presented with a list of current operator-specified FAFs and enters the desired changes to it. Step 14
At 2, the operator is presented with the current maximum number of FMVs and enters the desired changes to it. In step 143, a check is made to determine if the entered number falls within an acceptable range and branch accordingly. The allowable range is 0 to 4. Since 4 is the number of drives in the library, it is used as the maximum value in the range, which simplifies programming. In alternative embodiments, the maximum value of the range can be another number X that matches the number of drives or maximum mountable volumes in the library, or any other number desired. If the number is outside the acceptable range, then the method returns to step 142 to allow the operator to enter another desired maximum number of FMVs. If the input number is within the allowable range, the process proceeds to step 144. At step 144, a branch is taken depending on whether the operator has entered 0 as the maximum number of FMVs. If so, then at step 145 all library operations associated with the use of FMVs and FAFs are disabled. Otherwise, proceed to step 146, at which point the OLMSC
B110 is updated with the new data. Next step 1
At 47 the request comes out of the routine.

The UPDATE routine updates the most frequently used files and volumes by F respectively.
Occasionally initiated by the system controller 17 to ensure that it is properly designated as AF and FMV. Next, the UPDATE routine will be described with reference to FIG. The UPDATE routine will return to step 1 during the relative inactivity of the library.
Start with an interruption of the console 11 at 50. Step 1
At 51, the data in the optical disk map 115 is examined to determine a new set of FMVs ranked by mounting frequency up to the maximum number that can be selected. Usually, the maximum number of FMVs is automatically selected. However, under certain conditions (libraries with few volumes, libraries with most volumes having equal mounts, etc.), the maximum number may not be reached. This decision precludes selecting both volumes on the optical disc as FMVs because drive 4 only allows access to one volume (ie, one side) of the optical disc at a time. In an alternative embodiment where each optical disc is a single volume (ie, drive 4 can access either side of the disc mounted therein), such exclusion is not required.

At step 152, the data in the file usage database 116 is consulted to determine a new set of FAFs. Designated by the operator (DESIGNNA
The FAF (specified in the TE routine) is initially maintained as such. FA specified by the operator
If F exceeds the capacity of FMV, the UPDATE routine is aborted and an error message is returned (not shown). Then, another FAF is automatically selected according to its access count until the amount of data in the FAF reaches or is close to the FMV capacity. Step 153
MO to mount the most FMVs on drive 4
The UNT routine is called. In step 154, F
Non-FM with most FAFs not yet stored in MV
MOUNT to mount V in another drive 4
The routine is called. F specified by the operator
AFs are considered to be accessed more frequently than any automatically selected FAF, and automatically selected FAFs are ranked by their respective access counts.
Then, in step 155, all FAFs on such non-FMVs are transferred to the most FMVs, and in step 156,
By calling the DEMOUNT routine the non-FMV
Will be demounted.

At step 157, the process branches depending on the remaining capacity of the most FMV. If sufficient capacity remains in the most FMV, then return to step 154 to mount the non-FMV with the most remaining FMV not stored in the FMV. Eventually, the capacity of the most FMV is approached, and the process branches from step 157 to step 158 to call the DEMOUNT routine to demount the most FMV. Step 159 also branches depending on the situation of FMV.
If there is another FMV with sufficient residual capacity, return to step 153 and call the MOUNT routine to mount the next most FMV in drive 4. The process continues until the capacity of the last FMV is approached, and when approaching, the process branches from step 159 to step 160 and the UPD.
The ATE routine ends.

Although the present invention has been disclosed in an optical disk library environment, similar considerations apply the present invention to other types of libraries as well. In addition, many changes can be made to the library, such as the number of drives and storage cells. For example, in one alternative embodiment,
Library 1 has 32 storage cells 3 and 2 drives 4
Equipped with. The system controller 17 is placed outside the housing to reduce the size of the housing. Library 1
The other functions of are basically unchanged. Also, the usage frequency statistics can be based on parameters other than those described herein.

[0061]

According to the present invention, the most frequently used storage medium is most likely to be mounted, and the hit rate is improved.

[Brief description of drawings]

FIG. 1 is a front cutaway perspective view of an optical disk library of the present invention.

FIG. 2 is the same view as FIG. 1, except that the console panel is opened outward and the fan is removed.

3 is a rear cutaway perspective view of the optical disc library of FIGS. 1 and 2. FIG.

FIG. 4 is an enlarged view of the robotic picker and gripper of FIG.

FIG. 5 is a schematic diagram of the hardware of the optical disc library of FIGS.

FIG. 6 is a schematic configuration diagram of a system control device of the optical disk library of FIGS.

FIG. 7 is a schematic configuration diagram of an internal data structure created during initialization.

FIG. 8 is a system controller of an optical disk library according to the present invention which receives a network request received by its upper interface on its lower interface by SCS.
6 is a flowchart showing an operation when converting to an I command packet.

9 is a flow diagram illustrating the operation of FIG. 8 for a representative IFS entry point.

FIG. 10 is a flowchart of a MOUNT (mount) routine.

FIG. 11 is a flow chart of a DEMOUNT (demount) routine.

FIG. 12 is a flow chart of a DESIGNATE routine.

FIG. 13 is a flow chart of an UPDATE (update) routine.

[Explanation of symbols]

 1 Optical Disk Library 2 Housing 3 Optical Disk Storage Cell 4 Internal Optical Disk Drive 5 Robotic Picker 6 Gripper 10 External Optical Disk Drive 11 Console 12 Keyboard 17 System Controller 30 System Processor

 ─────────────────────────────────────────────────── ─── Continuation of the front page (72) Inventor Daniel James Winarski 647 South Woodstock Drive, Tucson, Arizona, USA 85710 (56) Reference JP-A-61-163429 (JP, A) Kai 61-262918 (JP, A) JP-A 63-244286 (JP, A) Actual Kaihei 2-273370 (JP, U)

Claims (9)

[Claims] 1. A storage device of a first hierarchical level and a storage device of a second hierarchical level, wherein the data stored in the storage devices of the first hierarchical level and the second hierarchical level are accessible as files and plural A method for improving a hit rate in a data storage device hierarchical structure such as transferable between the storage devices of the first hierarchical level and the second hierarchical level as a volume having files of the second hierarchical level. Transferring the volume from the storage device of the first tier level to the storage device of the first tier level, accessing files in the storage device of the first tier level, tracking the number of times each file is accessed, From the storage device at the second hierarchical level to the first
Recording the number of times a volume is transferred to a storage device at a tier level, determining which files are frequently accessed and which volumes are frequently transferred, and transferring the frequently accessed files to the frequently transferred volumes. Including the method. 2. The storage device of the first tier level has a storage capacity of X volumes, and the frequently accessed files are transferred to X or less frequently transferred volumes. The method of claim 1. 3. The data stored in the first hierarchical level storage device is data in a data storage medium in a peripheral storage device data access device of an automatic storage device library, and the second hierarchical level storage device. The method of claim 1, wherein the data stored in the device is data in a storage medium stored in a storage cell of the automatic storage library. 4. A first tier level storage device and a second tier level storage device, wherein data is stored in a format such that the data is accessible as a file and can be transferred to each other as a volume having a plurality of files. A storage device, means for transferring a volume from the storage device of the second hierarchical level to the storage device of the first hierarchical level, means for accessing a file in the storage device of the first hierarchical level, and the access means. Means for recording access of each file and transfer of each volume from the storage device of the second hierarchical level to the storage device of the first hierarchical level;
A data storage hierarchical structure comprising: means for determining a frequently accessed file and a frequently transferred volume in response to the tracking means; and means for transferring a frequently accessed file to a frequently transferred volume. 5. The storage device at the first hierarchical level has a storage capacity of X volumes, and the frequently accessed files are transferred to less than X frequently transferred volumes. 5. The data storage device hierarchical structure according to claim 4. 6. An automated storage device library for operating a storage medium including at least one volume of a plurality of files, the plurality of peripheral storage medium access devices each including one of said storage mediums. A plurality of storage cells capable of storing, moving the storage medium between the access device and the storage cell, mounting the storage medium, and mounting each volume within one of the access devices. Means for counting the number of times each file is accessed by the access device and the number of times each volume is mounted by the mount means in one of the access devices; and accessing in response to the counting means. How to determine which files and which volumes are mounted frequently Automatic storage library and means for transferring the high frequently mounted volume high file. 7. The peripheral storage medium access device has a total mountable capacity of X volumes, and the frequently accessed files are transferred to X or less frequently mounted volumes. Claim 6
Automatic storage library. 8. The peripheral storage medium access device has a total mountable capacity of X volumes, and the library further includes means for mounting X or less volumes that are frequently mounted at initialization. The automatic storage device library according to claim 6. 9. The automatic storage library of claim 6, further comprising means for allowing an operator to designate a particular file as the most frequently accessed file.